Time delay based failure-robust direction of arrival estimation

There is an increasing interest in autonomous and unattended sensory systems, including applications in smart meeting rooms and surveillance of large areas. In these systems, direction of arrival (DOA) estimation has an important role. Although several time delay based algorithms have been proposed, little attention has been given to hardware failures and malfunctions. These are, however, an important consideration when smart arrays and intelligent sensor systems are developed. In this paper, we address this problem by modifying a previously proposed single-failure detector to suit a multiple failure case. The performance of the modified method is tested both with simulations and speech data recorded using a microphone array in a meeting room. Results indicate that the proposed method can detect failures effectively. The detector is also combined with DOA estimators. Results show that this combination is robust against failures and the detector provides an efficient tool for array self-diagnosis.